Paragraph MCP
MCP server that connects AI agents to your Paragraph publication, enabling post management, subscriber analytics, and search across posts and publications.
README
@paragraph-com/mcp
MCP (Model Context Protocol) server for Paragraph — expose your publication to AI agents in Claude Desktop, Cursor, VS Code Copilot, and any MCP-compatible client.
Remote server (recommended)
Use the hosted server at mcp.paragraph.com — no installation or API key management required. You'll authenticate through your Paragraph account in the browser.
Claude Code
claude mcp add paragraph --transport http https://mcp.paragraph.com/mcp
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"paragraph": {
"url": "https://mcp.paragraph.com/mcp"
}
}
}
Cursor
Add to .cursor/mcp.json:
{
"mcpServers": {
"paragraph": {
"url": "https://mcp.paragraph.com/mcp"
}
}
}
VS Code
Add to .vscode/mcp.json:
{
"servers": {
"paragraph": {
"url": "https://mcp.paragraph.com/mcp"
}
}
}
Local server
If you prefer to run the server locally, you can use npx. Requires Node.js 18+.
npx @paragraph-com/mcp
Authentication (local only)
The remote server handles auth automatically. For local usage, set your API key via environment variable:
PARAGRAPH_API_KEY=your-key npx @paragraph-com/mcp
Or log in with the Paragraph CLI — the MCP server shares the same config:
npx @paragraph-com/cli login
npx @paragraph-com/mcp
Local setup by client
<details> <summary>Claude Code</summary>
claude mcp add paragraph -- npx @paragraph-com/mcp
</details>
<details> <summary>Claude Desktop</summary>
Add to claude_desktop_config.json:
{
"mcpServers": {
"paragraph": {
"command": "npx",
"args": ["@paragraph-com/mcp"],
"env": {
"PARAGRAPH_API_KEY": "your-key"
}
}
}
}
</details>
<details> <summary>Cursor / VS Code</summary>
Add to .cursor/mcp.json or .vscode/mcp.json:
{
"servers": {
"paragraph": {
"command": "npx",
"args": ["@paragraph-com/mcp"],
"env": {
"PARAGRAPH_API_KEY": "your-key"
}
}
}
}
</details>
Streamable HTTP (local)
npx @paragraph-com/mcp --http --port 3100
HTTP mode binds to
127.0.0.1(localhost only) and accepts POST requests. A health check is available atGET /health.
Available Tools
Posts
- get-post — Get a post by ID or slugs
- list-posts — List posts from a publication or your own posts
- create-post — Create a new post (markdown)
- update-post — Update an existing post
- delete-post — Delete a post
- send-test-email — Send a test newsletter for a draft
Publications
- get-publication — Get publication metadata by ID, slug, or domain
Subscribers
- list-subscribers — List your subscribers
- get-subscriber-count — Get subscriber count
- add-subscriber — Add a subscriber by email or wallet
Users
- get-user — Get user profile by ID or wallet
Coins
- get-coin — Get coin metadata or list popular coins
- list-coin-holders — Get coin holders
Search
- search-posts — Search posts
- search-blogs — Search publications
- search-coins — Search coins
Feed
- get-feed — Get the curated platform feed
Me
- get-me — Get your authenticated publication info
Analytics
- analytics-query — Run a read-only SQL query against your publication's analytics (open rates, CTR, subscriber counts, top posts, etc.)
- analytics-schema — List available tables and columns in the analytics schema
Toolset Filtering
Only expose the tools your agent needs:
npx @paragraph-com/mcp --toolsets posts,search
Available toolsets: posts, publications, subscribers, users, coins, search, feed, me, analytics
Examples
Search and summarize posts on a topic
Prompt: "Search for posts about onchain governance and give me a summary of the top results"
The MCP server calls search-posts with the query. Claude receives the matching posts and synthesizes a summary of the key themes, authors, and publications.
Draft and publish a blog post
Prompt: "Write a post titled 'Weekly Update #12' about our new token-gating feature, then publish it"
The server calls create-post with the title and generated markdown content, creating it as a draft. After you confirm, it calls update-post to set the status to published, making it live on your publication.
Get subscriber analytics
Prompt: "How many subscribers does my publication have? Show me the most recent ones."
The server calls get-me to identify your publication, then get-subscriber-count for the total, and list-subscribers to return the latest subscribers with their details.
Answer questions about newsletter performance
Prompt: "What's my average open rate over the last 30 days, and which posts performed best?"
The server calls analytics-query with a scoped SQL query against post_analytics_summary, returning open rates, CTR, and views for each post. Claude summarizes the trend and highlights the top performers.
Explore the platform feed
Prompt: "What's trending on Paragraph right now?"
The server calls get-feed to retrieve the curated platform feed and Claude summarizes the top posts, their authors, and topics.
Manage post lifecycle
Prompt: "Show me my drafts and send a test email for the most recent one"
The server calls list-posts with status: draft to retrieve your drafts, then send-test-email with the latest draft's ID so you can preview the newsletter in your inbox before sending it to subscribers.
Support
- Issues & bugs: GitHub Issues
- Email: support@paragraph.com
- Website: paragraph.com
Releasing
npm run release patch # 1.0.0 → 1.0.1
npm run release minor # 1.0.0 → 1.1.0
npm run release major # 1.0.0 → 2.0.0
Builds, tests, publishes to npm, deploys the Cloudflare Worker at mcp.paragraph.com, and creates a GitHub release. The script will prompt for any missing credentials.
Privacy Policy
This MCP server connects to the Paragraph API on your behalf. See the Paragraph Privacy Policy for details on data collection, usage, storage, and your rights.
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。